angeluriot/chess_games

Chess games数据集是一个包含1400万局高水平国际象棋比赛的数据集，涵盖了1600年至2024年间的12亿步棋。数据集的平均ELO为2388，每局平均步数为84步。数据集主要用于训练机器学习模型，也可用于数据分析和可视化。每局比赛的数据结构包括比赛日期、白方和黑方的ELO评分、比赛结束方式、获胜者、SAN和UCI格式的棋步、自定义棋步表示以及比赛来源。

The dataset is a collection of high-level chess games, primarily intended for training machine learning models. It contains 14 million games, totaling 1.2 billion moves, with games spanning from 1600 to 2024, most of which are recent. The average ELO rating of the players is 2388, with an average of 84 moves per game and a maximum of 692 moves. Games typically end in resignation or draw agreements. Each record in the dataset includes the date of the game, ELO ratings of the white and black players, the way the game ended, the winner, moves in Standard Algebraic Notation (SAN), Universal Chess Interface (UCI), and custom notation, as well as the source of the game. The sources of the dataset include Lichess Elite Database, Lumbras Giga Base, PGN Mentor, etc., totaling 14,188,454 games.